| H A D | topics.py | 20 'calls': u'\nCalls\n*****\n\nA call calls a callable object (e.g., a *function*) with a possibly\nempty series of *arguments*:\n\n call ::= primary "(" [argument_list [","]\n | expression genexpr_for] ")"\n argument_list ::= positional_arguments ["," keyword_arguments]\n ["," "*" expression] ["," keyword_arguments]\n ["," "**" expression]\n | keyword_arguments ["," "*" expression]\n ["," "**" expression]\n | "*" expression ["," keyword_arguments] ["," "**" expression]\n | "**" expression\n positional_arguments ::= expression ("," expression)*\n keyword_arguments ::= keyword_item ("," keyword_item)*\n keyword_item ::= identifier "=" expression\n\nA trailing comma may be present after the positional and keyword\narguments but does not affect the semantics.\n\nThe primary must evaluate to a callable object (user-defined\nfunctions, built-in functions, methods of built-in objects, class\nobjects, methods of class instances, and certain class instances\nthemselves are callable; extensions may define additional callable\nobject types). All argument expressions are evaluated before the call\nis attempted. Please refer to section Function definitions for the\nsyntax of formal *parameter* lists.\n\nIf keyword arguments are present, they are first converted to\npositional arguments, as follows. First, a list of unfilled slots is\ncreated for the formal parameters. If there are N positional\narguments, they are placed in the first N slots. Next, for each\nkeyword argument, the identifier is used to determine the\ncorresponding slot (if the identifier is the same as the first formal\nparameter name, the first slot is used, and so on). If the slot is\nalready filled, a "TypeError" exception is raised. Otherwise, the\nvalue of the argument is placed in the slot, filling it (even if the\nexpression is "None", it fills the slot). When all arguments have\nbeen processed, the slots that are still unfilled are filled with the\ncorresponding default value from the function definition. (Default\nvalues are calculated, once, when the function is defined; thus, a\nmutable object such as a list or dictionary used as default value will\nbe shared by all calls that don\'t specify an argument value for the\ncorresponding slot; this should usually be avoided.) If there are any\nunfilled slots for which no default value is specified, a "TypeError"\nexception is raised. Otherwise, the list of filled slots is used as\nthe argument list for the call.\n\n**CPython implementation detail:** An implementation may provide\nbuilt-in functions whose positional parameters do not have names, even\nif they are \'named\' for the purpose of documentation, and which\ntherefore cannot be supplied by keyword. In CPython, this is the case\nfor functions implemented in C that use "PyArg_ParseTuple()" to parse\ntheir arguments.\n\nIf there are more positional arguments than there are formal parameter\nslots, a "TypeError" exception is raised, unless a formal parameter\nusing the syntax "*identifier" is present; in this case, that formal\nparameter receives a tuple containing the excess positional arguments\n(or an empty tuple if there were no excess positional arguments).\n\nIf any keyword argument does not correspond to a formal parameter\nname, a "TypeError" exception is raised, unless a formal parameter\nusing the syntax "**identifier" is present; in this case, that formal\nparameter receives a dictionary containing the excess keyword\narguments (using the keywords as keys and the argument values as\ncorresponding values), or a (new) empty dictionary if there were no\nexcess keyword arguments.\n\nIf the syntax "*expression" appears in the function call, "expression"\nmust evaluate to an iterable. Elements from this iterable are treated\nas if they were additional positional arguments; if there are\npositional arguments *x1*, ..., *xN*, and "expression" evaluates to a\nsequence *y1*, ..., *yM*, this is equivalent to a call with M+N\npositional arguments *x1*, ..., *xN*, *y1*, ..., *yM*.\n\nA consequence of this is that although the "*expression" syntax may\nappear *after* some keyword arguments, it is processed *before* the\nkeyword arguments (and the "**expression" argument, if any -- see\nbelow). So:\n\n >>> def f(a, b):\n ... print a, b\n ...\n >>> f(b=1, *(2,))\n 2 1\n >>> f(a=1, *(2,))\n Traceback (most recent call last):\n File "<stdin>", line 1, in ?\n TypeError: f() got multiple values for keyword argument \'a\'\n >>> f(1, *(2,))\n 1 2\n\nIt is unusual for both keyword arguments and the "*expression" syntax\nto be used in the same call, so in practice this confusion does not\narise.\n\nIf the syntax "**expression" appears in the function call,\n"expression" must evaluate to a mapping, the contents of which are\ntreated as additional keyword arguments. In the case of a keyword\nappearing in both "expression" and as an explicit keyword argument, a\n"TypeError" exception is raised.\n\nFormal parameters using the syntax "*identifier" or "**identifier"\ncannot be used as positional argument slots or as keyword argument\nnames. Formal parameters using the syntax "(sublist)" cannot be used\nas keyword argument names; the outermost sublist corresponds to a\nsingle unnamed argument slot, and the argument value is assigned to\nthe sublist using the usual tuple assignment rules after all other\nparameter processing is done.\n\nA call always returns some value, possibly "None", unless it raises an\nexception. How this value is computed depends on the type of the\ncallable object.\n\nIf it is---\n\na user-defined function:\n The code block for the function is executed, passing it the\n argument list. The first thing the code block will do is bind the\n formal parameters to the arguments; this is described in section\n Function definitions. When the code block executes a "return"\n statement, this specifies the return value of the function call.\n\na built-in function or method:\n The result is up to the interpreter; see Built-in Functions for the\n descriptions of built-in functions and methods.\n\na class object:\n A new instance of that class is returned.\n\na class instance method:\n The corresponding user-defined function is called, with an argument\n list that is one longer than the argument list of the call: the\n instance becomes the first argument.\n\na class instance:\n The class must define a "__call__()" method; the effect is then the\n same as if that method was called.\n',
27 'customization': u'\nBasic customization\n*******************\n\nobject.__new__(cls[, ...])\n\n Called to create a new instance of class *cls*. "__new__()" is a\n static method (special-cased so you need not declare it as such)\n that takes the class of which an instance was requested as its\n first argument. The remaining arguments are those passed to the\n object constructor expression (the call to the class). The return\n value of "__new__()" should be the new object instance (usually an\n instance of *cls*).\n\n Typical implementations create a new instance of the class by\n invoking the superclass\'s "__new__()" method using\n "super(currentclass, cls).__new__(cls[, ...])" with appropriate\n arguments and then modifying the newly-created instance as\n necessary before returning it.\n\n If "__new__()" returns an instance of *cls*, then the new\n instance\'s "__init__()" method will be invoked like\n "__init__(self[, ...])", where *self* is the new instance and the\n remaining arguments are the same as were passed to "__new__()".\n\n If "__new__()" does not return an instance of *cls*, then the new\n instance\'s "__init__()" method will not be invoked.\n\n "__new__()" is intended mainly to allow subclasses of immutable\n types (like int, str, or tuple) to customize instance creation. It\n is also commonly overridden in custom metaclasses in order to\n customize class creation.\n\nobject.__init__(self[, ...])\n\n Called after the instance has been created (by "__new__()"), but\n before it is returned to the caller. The arguments are those\n passed to the class constructor expression. If a base class has an\n "__init__()" method, the derived class\'s "__init__()" method, if\n any, must explicitly call it to ensure proper initialization of the\n base class part of the instance; for example:\n "BaseClass.__init__(self, [args...])".\n\n Because "__new__()" and "__init__()" work together in constructing\n objects ("__new__()" to create it, and "__init__()" to customise\n it), no non-"None" value may be returned by "__init__()"; doing so\n will cause a "TypeError" to be raised at runtime.\n\nobject.__del__(self)\n\n Called when the instance is about to be destroyed. This is also\n called a destructor. If a base class has a "__del__()" method, the\n derived class\'s "__del__()" method, if any, must explicitly call it\n to ensure proper deletion of the base class part of the instance.\n Note that it is possible (though not recommended!) for the\n "__del__()" method to postpone destruction of the instance by\n creating a new reference to it. It may then be called at a later\n time when this new reference is deleted. It is not guaranteed that\n "__del__()" methods are called for objects that still exist when\n the interpreter exits.\n\n Note: "del x" doesn\'t directly call "x.__del__()" --- the former\n decrements the reference count for "x" by one, and the latter is\n only called when "x"\'s reference count reaches zero. Some common\n situations that may prevent the reference count of an object from\n going to zero include: circular references between objects (e.g.,\n a doubly-linked list or a tree data structure with parent and\n child pointers); a reference to the object on the stack frame of\n a function that caught an exception (the traceback stored in\n "sys.exc_traceback" keeps the stack frame alive); or a reference\n to the object on the stack frame that raised an unhandled\n exception in interactive mode (the traceback stored in\n "sys.last_traceback" keeps the stack frame alive). The first\n situation can only be remedied by explicitly breaking the cycles;\n the latter two situations can be resolved by storing "None" in\n "sys.exc_traceback" or "sys.last_traceback". Circular references\n which are garbage are detected when the option cycle detector is\n enabled (it\'s on by default), but can only be cleaned up if there\n are no Python-level "__del__()" methods involved. Refer to the\n documentation for the "gc" module for more information about how\n "__del__()" methods are handled by the cycle detector,\n particularly the description of the "garbage" value.\n\n Warning: Due to the precarious circumstances under which\n "__del__()" methods are invoked, exceptions that occur during\n their execution are ignored, and a warning is printed to\n "sys.stderr" instead. Also, when "__del__()" is invoked in\n response to a module being deleted (e.g., when execution of the\n program is done), other globals referenced by the "__del__()"\n method may already have been deleted or in the process of being\n torn down (e.g. the import machinery shutting down). For this\n reason, "__del__()" methods should do the absolute minimum needed\n to maintain external invariants. Starting with version 1.5,\n Python guarantees that globals whose name begins with a single\n underscore are deleted from their module before other globals are\n deleted; if no other references to such globals exist, this may\n help in assuring that imported modules are still available at the\n time when the "__del__()" method is called.\n\n See also the "-R" command-line option.\n\nobject.__repr__(self)\n\n Called by the "repr()" built-in function and by string conversions\n (reverse quotes) to compute the "official" string representation of\n an object. If at all possible, this should look like a valid\n Python expression that could be used to recreate an object with the\n same value (given an appropriate environment). If this is not\n possible, a string of the form "<...some useful description...>"\n should be returned. The return value must be a string object. If a\n class defines "__repr__()" but not "__str__()", then "__repr__()"\n is also used when an "informal" string representation of instances\n of that class is required.\n\n This is typically used for debugging, so it is important that the\n representation is information-rich and unambiguous.\n\nobject.__str__(self)\n\n Called by the "str()" built-in function and by the "print"\n statement to compute the "informal" string representation of an\n object. This differs from "__repr__()" in that it does not have to\n be a valid Python expression: a more convenient or concise\n representation may be used instead. The return value must be a\n string object.\n\nobject.__lt__(self, other)\nobject.__le__(self, other)\nobject.__eq__(self, other)\nobject.__ne__(self, other)\nobject.__gt__(self, other)\nobject.__ge__(self, other)\n\n New in version 2.1.\n\n These are the so-called "rich comparison" methods, and are called\n for comparison operators in preference to "__cmp__()" below. The\n correspondence between operator symbols and method names is as\n follows: "x<y" calls "x.__lt__(y)", "x<=y" calls "x.__le__(y)",\n "x==y" calls "x.__eq__(y)", "x!=y" and "x<>y" call "x.__ne__(y)",\n "x>y" calls "x.__gt__(y)", and "x>=y" calls "x.__ge__(y)".\n\n A rich comparison method may return the singleton "NotImplemented"\n if it does not implement the operation for a given pair of\n arguments. By convention, "False" and "True" are returned for a\n successful comparison. However, these methods can return any value,\n so if the comparison operator is used in a Boolean context (e.g.,\n in the condition of an "if" statement), Python will call "bool()"\n on the value to determine if the result is true or false.\n\n There are no implied relationships among the comparison operators.\n The truth of "x==y" does not imply that "x!=y" is false.\n Accordingly, when defining "__eq__()", one should also define\n "__ne__()" so that the operators will behave as expected. See the\n paragraph on "__hash__()" for some important notes on creating\n *hashable* objects which support custom comparison operations and\n are usable as dictionary keys.\n\n There are no swapped-argument versions of these methods (to be used\n when the left argument does not support the operation but the right\n argument does); rather, "__lt__()" and "__gt__()" are each other\'s\n reflection, "__le__()" and "__ge__()" are each other\'s reflection,\n and "__eq__()" and "__ne__()" are their own reflection.\n\n Arguments to rich comparison methods are never coerced.\n\n To automatically generate ordering operations from a single root\n operation, see "functools.total_ordering()".\n\nobject.__cmp__(self, other)\n\n Called by comparison operations if rich comparison (see above) is\n not defined. Should return a negative integer if "self < other",\n zero if "self == other", a positive integer if "self > other". If\n no "__cmp__()", "__eq__()" or "__ne__()" operation is defined,\n class instances are compared by object identity ("address"). See\n also the description of "__hash__()" for some important notes on\n creating *hashable* objects which support custom comparison\n operations and are usable as dictionary keys. (Note: the\n restriction that exceptions are not propagated by "__cmp__()" has\n been removed since Python 1.5.)\n\nobject.__rcmp__(self, other)\n\n Changed in version 2.1: No longer supported.\n\nobject.__hash__(self)\n\n Called by built-in function "hash()" and for operations on members\n of hashed collections including "set", "frozenset", and "dict".\n "__hash__()" should return an integer. The only required property\n is that objects which compare equal have the same hash value; it is\n advised to somehow mix together (e.g. using exclusive or) the hash\n values for the components of the object that also play a part in\n comparison of objects.\n\n If a class does not define a "__cmp__()" or "__eq__()" method it\n should not define a "__hash__()" operation either; if it defines\n "__cmp__()" or "__eq__()" but not "__hash__()", its instances will\n not be usable in hashed collections. If a class defines mutable\n objects and implements a "__cmp__()" or "__eq__()" method, it\n should not implement "__hash__()", since hashable collection\n implementations require that an object\'s hash value is immutable\n (if the object\'s hash value changes, it will be in the wrong hash\n bucket).\n\n User-defined classes have "__cmp__()" and "__hash__()" methods by\n default; with them, all objects compare unequal (except with\n themselves) and "x.__hash__()" returns a result derived from\n "id(x)".\n\n Classes which inherit a "__hash__()" method from a parent class but\n change the meaning of "__cmp__()" or "__eq__()" such that the hash\n value returned is no longer appropriate (e.g. by switching to a\n value-based concept of equality instead of the default identity\n based equality) can explicitly flag themselves as being unhashable\n by setting "__hash__ = None" in the class definition. Doing so\n means that not only will instances of the class raise an\n appropriate "TypeError" when a program attempts to retrieve their\n hash value, but they will also be correctly identified as\n unhashable when checking "isinstance(obj, collections.Hashable)"\n (unlike classes which define their own "__hash__()" to explicitly\n raise "TypeError").\n\n Changed in version 2.5: "__hash__()" may now also return a long\n integer object; the 32-bit integer is then derived from the hash of\n that object.\n\n Changed in version 2.6: "__hash__" may now be set to "None" to\n explicitly flag instances of a class as unhashable.\n\nobject.__nonzero__(self)\n\n Called to implement truth value testing and the built-in operation\n "bool()"; should return "False" or "True", or their integer\n equivalents "0" or "1". When this method is not defined,\n "__len__()" is called, if it is defined, and the object is\n considered true if its result is nonzero. If a class defines\n neither "__len__()" nor "__nonzero__()", all its instances are\n considered true.\n\nobject.__unicode__(self)\n\n Called to implement "unicode()" built-in; should return a Unicode\n object. When this method is not defined, string conversion is\n attempted, and the result of string conversion is converted to\n Unicode using the system default encoding.\n', namespace
59 'raise': u'\nThe "raise" statement\n*********************\n\n raise_stmt ::= "raise" [expression ["," expression ["," expression]]]\n\nIf no expressions are present, "raise" re-raises the last exception\nthat was active in the current scope. If no exception is active in\nthe current scope, a "TypeError" exception is raised indicating that\nthis is an error (if running under IDLE, a "Queue.Empty" exception is\nraised instead).\n\nOtherwise, "raise" evaluates the expressions to get three objects,\nusing "None" as the value of omitted expressions. The first two\nobjects are used to determine the *type* and *value* of the exception.\n\nIf the first object is an instance, the type of the exception is the\nclass of the instance, the instance itself is the value, and the\nsecond object must be "None".\n\nIf the first object is a class, it becomes the type of the exception.\nThe second object is used to determine the exception value: If it is\nan instance of the class, the instance becomes the exception value. If\nthe second object is a tuple, it is used as the argument list for the\nclass constructor; if it is "None", an empty argument list is used,\nand any other object is treated as a single argument to the\nconstructor. The instance so created by calling the constructor is\nused as the exception value.\n\nIf a third object is present and not "None", it must be a traceback\nobject (see section The standard type hierarchy), and it is\nsubstituted instead of the current location as the place where the\nexception occurred. If the third object is present and not a\ntraceback object or "None", a "TypeError" exception is raised. The\nthree-expression form of "raise" is useful to re-raise an exception\ntransparently in an except clause, but "raise" with no expressions\nshould be preferred if the exception to be re-raised was the most\nrecently active exception in the current scope.\n\nAdditional information on exceptions can be found in section\nExceptions, and information about handling exceptions is in section\nThe try statement.\n',
65 'specialnames': u'\nSpecial method names\n********************\n\nA class can implement certain operations that are invoked by special\nsyntax (such as arithmetic operations or subscripting and slicing) by\ndefining methods with special names. This is Python\'s approach to\n*operator overloading*, allowing classes to define their own behavior\nwith respect to language operators. For instance, if a class defines\na method named "__getitem__()", and "x" is an instance of this class,\nthen "x[i]" is roughly equivalent to "x.__getitem__(i)" for old-style\nclasses and "type(x).__getitem__(x, i)" for new-style classes. Except\nwhere mentioned, attempts to execute an operation raise an exception\nwhen no appropriate method is defined (typically "AttributeError" or\n"TypeError").\n\nWhen implementing a class that emulates any built-in type, it is\nimportant that the emulation only be implemented to the degree that it\nmakes sense for the object being modelled. For example, some\nsequences may work well with retrieval of individual elements, but\nextracting a slice may not make sense. (One example of this is the\n"NodeList" interface in the W3C\'s Document Object Model.)\n\n\nBasic customization\n===================\n\nobject.__new__(cls[, ...])\n\n Called to create a new instance of class *cls*. "__new__()" is a\n static method (special-cased so you need not declare it as such)\n that takes the class of which an instance was requested as its\n first argument. The remaining arguments are those passed to the\n object constructor expression (the call to the class). The return\n value of "__new__()" should be the new object instance (usually an\n instance of *cls*).\n\n Typical implementations create a new instance of the class by\n invoking the superclass\'s "__new__()" method using\n "super(currentclass, cls).__new__(cls[, ...])" with appropriate\n arguments and then modifying the newly-created instance as\n necessary before returning it.\n\n If "__new__()" returns an instance of *cls*, then the new\n instance\'s "__init__()" method will be invoked like\n "__init__(self[, ...])", where *self* is the new instance and the\n remaining arguments are the same as were passed to "__new__()".\n\n If "__new__()" does not return an instance of *cls*, then the new\n instance\'s "__init__()" method will not be invoked.\n\n "__new__()" is intended mainly to allow subclasses of immutable\n types (like int, str, or tuple) to customize instance creation. It\n is also commonly overridden in custom metaclasses in order to\n customize class creation.\n\nobject.__init__(self[, ...])\n\n Called after the instance has been created (by "__new__()"), but\n before it is returned to the caller. The arguments are those\n passed to the class constructor expression. If a base class has an\n "__init__()" method, the derived class\'s "__init__()" method, if\n any, must explicitly call it to ensure proper initialization of the\n base class part of the instance; for example:\n "BaseClass.__init__(self, [args...])".\n\n Because "__new__()" and "__init__()" work together in constructing\n objects ("__new__()" to create it, and "__init__()" to customise\n it), no non-"None" value may be returned by "__init__()"; doing so\n will cause a "TypeError" to be raised at runtime.\n\nobject.__del__(self)\n\n Called when the instance is about to be destroyed. This is also\n called a destructor. If a base class has a "__del__()" method, the\n derived class\'s "__del__()" method, if any, must explicitly call it\n to ensure proper deletion of the base class part of the instance.\n Note that it is possible (though not recommended!) for the\n "__del__()" method to postpone destruction of the instance by\n creating a new reference to it. It may then be called at a later\n time when this new reference is deleted. It is not guaranteed that\n "__del__()" methods are called for objects that still exist when\n the interpreter exits.\n\n Note: "del x" doesn\'t directly call "x.__del__()" --- the former\n decrements the reference count for "x" by one, and the latter is\n only called when "x"\'s reference count reaches zero. Some common\n situations that may prevent the reference count of an object from\n going to zero include: circular references between objects (e.g.,\n a doubly-linked list or a tree data structure with parent and\n child pointers); a reference to the object on the stack frame of\n a function that caught an exception (the traceback stored in\n "sys.exc_traceback" keeps the stack frame alive); or a reference\n to the object on the stack frame that raised an unhandled\n exception in interactive mode (the traceback stored in\n "sys.last_traceback" keeps the stack frame alive). The first\n situation can only be remedied by explicitly breaking the cycles;\n the latter two situations can be resolved by storing "None" in\n "sys.exc_traceback" or "sys.last_traceback". Circular references\n which are garbage are detected when the option cycle detector is\n enabled (it\'s on by default), but can only be cleaned up if there\n are no Python-level "__del__()" methods involved. Refer to the\n documentation for the "gc" module for more information about how\n "__del__()" methods are handled by the cycle detector,\n particularly the description of the "garbage" value.\n\n Warning: Due to the precarious circumstances under which\n "__del__()" methods are invoked, exceptions that occur during\n their execution are ignored, and a warning is printed to\n "sys.stderr" instead. Also, when "__del__()" is invoked in\n response to a module being deleted (e.g., when execution of the\n program is done), other globals referenced by the "__del__()"\n method may already have been deleted or in the process of being\n torn down (e.g. the import machinery shutting down). For this\n reason, "__del__()" methods should do the absolute minimum needed\n to maintain external invariants. Starting with version 1.5,\n Python guarantees that globals whose name begins with a single\n underscore are deleted from their module before other globals are\n deleted; if no other references to such globals exist, this may\n help in assuring that imported modules are still available at the\n time when the "__del__()" method is called.\n\n See also the "-R" command-line option.\n\nobject.__repr__(self)\n\n Called by the "repr()" built-in function and by string conversions\n (reverse quotes) to compute the "official" string representation of\n an object. If at all possible, this should look like a valid\n Python expression that could be used to recreate an object with the\n same value (given an appropriate environment). If this is not\n possible, a string of the form "<...some useful description...>"\n should be returned. The return value must be a string object. If a\n class defines "__repr__()" but not "__str__()", then "__repr__()"\n is also used when an "informal" string representation of instances\n of that class is required.\n\n This is typically used for debugging, so it is important that the\n representation is information-rich and unambiguous.\n\nobject.__str__(self)\n\n Called by the "str()" built-in function and by the "print"\n statement to compute the "informal" string representation of an\n object. This differs from "__repr__()" in that it does not have to\n be a valid Python expression: a more convenient or concise\n representation may be used instead. The return value must be a\n string object.\n\nobject.__lt__(self, other)\nobject.__le__(self, other)\nobject.__eq__(self, other)\nobject.__ne__(self, other)\nobject.__gt__(self, other)\nobject.__ge__(self, other)\n\n New in version 2.1.\n\n These are the so-called "rich comparison" methods, and are called\n for comparison operators in preference to "__cmp__()" below. The\n correspondence between operator symbols and method names is as\n follows: "x<y" calls "x.__lt__(y)", "x<=y" calls "x.__le__(y)",\n "x==y" calls "x.__eq__(y)", "x!=y" and "x<>y" call "x.__ne__(y)",\n "x>y" calls "x.__gt__(y)", and "x>=y" calls "x.__ge__(y)".\n\n A rich comparison method may return the singleton "NotImplemented"\n if it does not implement the operation for a given pair of\n arguments. By convention, "False" and "True" are returned for a\n successful comparison. However, these methods can return any value,\n so if the comparison operator is used in a Boolean context (e.g.,\n in the condition of an "if" statement), Python will call "bool()"\n on the value to determine if the result is true or false.\n\n There are no implied relationships among the comparison operators.\n The truth of "x==y" does not imply that "x!=y" is false.\n Accordingly, when defining "__eq__()", one should also define\n "__ne__()" so that the operators will behave as expected. See the\n paragraph on "__hash__()" for some important notes on creating\n *hashable* objects which support custom comparison operations and\n are usable as dictionary keys.\n\n There are no swapped-argument versions of these methods (to be used\n when the left argument does not support the operation but the right\n argument does); rather, "__lt__()" and "__gt__()" are each other\'s\n reflection, "__le__()" and "__ge__()" are each other\'s reflection,\n and "__eq__()" and "__ne__()" are their own reflection.\n\n Arguments to rich comparison methods are never coerced.\n\n To automatically generate ordering operations from a single root\n operation, see "functools.total_ordering()".\n\nobject.__cmp__(self, other)\n\n Called by comparison operations if rich comparison (see above) is\n not defined. Should return a negative integer if "self < other",\n zero if "self == other", a positive integer if "self > other". If\n no "__cmp__()", "__eq__()" or "__ne__()" operation is defined,\n class instances are compared by object identity ("address"). See\n also the description of "__hash__()" for some important notes on\n creating *hashable* objects which support custom comparison\n operations and are usable as dictionary keys. (Note: the\n restriction that exceptions are not propagated by "__cmp__()" has\n been removed since Python 1.5.)\n\nobject.__rcmp__(self, other)\n\n Changed in version 2.1: No longer supported.\n\nobject.__hash__(self)\n\n Called by built-in function "hash()" and for operations on members\n of hashed collections including "set", "frozenset", and "dict".\n "__hash__()" should return an integer. The only required property\n is that objects which compare equal have the same hash value; it is\n advised to somehow mix together (e.g. using exclusive or) the hash\n values for the components of the object that also play a part in\n comparison of objects.\n\n If a class does not define a "__cmp__()" or "__eq__()" method it\n should not define a "__hash__()" operation either; if it defines\n "__cmp__()" or "__eq__()" but not "__hash__()", its instances will\n not be usable in hashed collections. If a class defines mutable\n objects and implements a "__cmp__()" or "__eq__()" method, it\n should not implement "__hash__()", since hashable collection\n implementations require that an object\'s hash value is immutable\n (if the object\'s hash value changes, it will be in the wrong hash\n bucket).\n\n User-defined classes have "__cmp__()" and "__hash__()" methods by\n default; with them, all objects compare unequal (except with\n themselves) and "x.__hash__()" returns a result derived from\n "id(x)".\n\n Classes which inherit a "__hash__()" method from a parent class but\n change the meaning of "__cmp__()" or "__eq__()" such that the hash\n value returned is no longer appropriate (e.g. by switching to a\n value-based concept of equality instead of the default identity\n based equality) can explicitly flag themselves as being unhashable\n by setting "__hash__ = None" in the class definition. Doing so\n means that not only will instances of the class raise an\n appropriate "TypeError" when a program attempts to retrieve their\n hash value, but they will also be correctly identified as\n unhashable when checking "isinstance(obj, collections.Hashable)"\n (unlike classes which define their own "__hash__()" to explicitly\n raise "TypeError").\n\n Changed in version 2.5: "__hash__()" may now also return a long\n integer object; the 32-bit integer is then derived from the hash of\n that object.\n\n Changed in version 2.6: "__hash__" may now be set to "None" to\n explicitly flag instances of a class as unhashable.\n\nobject.__nonzero__(self)\n\n Called to implement truth value testing and the built-in operation\n "bool()"; should return "False" or "True", or their integer\n equivalents "0" or "1". When this method is not defined,\n "__len__()" is called, if it is defined, and the object is\n considered true if its result is nonzero. If a class defines\n neither "__len__()" nor "__nonzero__()", all its instances are\n considered true.\n\nobject.__unicode__(self)\n\n Called to implement "unicode()" built-in; should return a Unicode\n object. When this method is not defined, string conversion is\n attempted, and the result of string conversion is converted to\n Unicode using the system default encoding.\n\n\nCustomizing attribute access\n============================\n\nThe following methods can be defined to customize the meaning of\nattribute access (use of, assignment to, or deletion of "x.name") for\nclass instances.\n\nobject.__getattr__(self, name)\n\n Called when an attribute lookup has not found the attribute in the\n usual places (i.e. it is not an instance attribute nor is it found\n in the class tree for "self"). "name" is the attribute name. This\n method should return the (computed) attribute value or raise an\n "AttributeError" exception.\n\n Note that if the attribute is found through the normal mechanism,\n "__getattr__()" is not called. (This is an intentional asymmetry\n between "__getattr__()" and "__setattr__()".) This is done both for\n efficiency reasons and because otherwise "__getattr__()" would have\n no way to access other attributes of the instance. Note that at\n least for instance variables, you can fake total control by not\n inserting any values in the instance attribute dictionary (but\n instead inserting them in another object). See the\n "__getattribute__()" method below for a way to actually get total\n control in new-style classes.\n\nobject.__setattr__(self, name, value)\n\n Called when an attribute assignment is attempted. This is called\n instead of the normal mechanism (i.e. store the value in the\n instance dictionary). *name* is the attribute name, *value* is the\n value to be assigned to it.\n\n If "__setattr__()" wants to assign to an instance attribute, it\n should not simply execute "self.name = value" --- this would cause\n a recursive call to itself. Instead, it should insert the value in\n the dictionary of instance attributes, e.g., "self.__dict__[name] =\n value". For new-style classes, rather than accessing the instance\n dictionary, it should call the base class method with the same\n name, for example, "object.__setattr__(self, name, value)".\n\nobject.__delattr__(self, name)\n\n Like "__setattr__()" but for attribute deletion instead of\n assignment. This should only be implemented if "del obj.name" is\n meaningful for the object.\n\n\nMore attribute access for new-style classes\n-------------------------------------------\n\nThe following methods only apply to new-style classes.\n\nobject.__getattribute__(self, name)\n\n Called unconditionally to implement attribute accesses for\n instances of the class. If the class also defines "__getattr__()",\n the latter will not be called unless "__getattribute__()" either\n calls it explicitly or raises an "AttributeError". This method\n should return the (computed) attribute value or raise an\n "AttributeError" exception. In order to avoid infinite recursion in\n this method, its implementation should always call the base class\n method with the same name to access any attributes it needs, for\n example, "object.__getattribute__(self, name)".\n\n Note: This method may still be bypassed when looking up special\n methods as the result of implicit invocation via language syntax\n or built-in functions. See Special method lookup for new-style\n classes.\n\n\nImplementing Descriptors\n------------------------\n\nThe following methods only apply when an instance of the class\ncontaining the method (a so-called *descriptor* class) appears in an\n*owner* class (the descriptor must be in either the owner\'s class\ndictionary or in the class dictionary for one of its parents). In the\nexamples below, "the attribute" refers to the attribute whose name is\nthe key of the property in the owner class\' "__dict__".\n\nobject.__get__(self, instance, owner)\n\n Called to get the attribute of the owner class (class attribute\n access) or of an instance of that class (instance attribute\n access). *owner* is always the owner class, while *instance* is the\n instance that the attribute was accessed through, or "None" when\n the attribute is accessed through the *owner*. This method should\n return the (computed) attribute value or raise an "AttributeError"\n exception.\n\nobject.__set__(self, instance, value)\n\n Called to set the attribute on an instance *instance* of the owner\n class to a new value, *value*.\n\nobject.__delete__(self, instance)\n\n Called to delete the attribute on an instance *instance* of the\n owner class.\n\n\nInvoking Descriptors\n--------------------\n\nIn general, a descriptor is an object attribute with "binding\nbehavior", one whose attribute access has been overridden by methods\nin the descriptor protocol: "__get__()", "__set__()", and\n"__delete__()". If any of those methods are defined for an object, it\nis said to be a descriptor.\n\nThe default behavior for attribute access is to get, set, or delete\nthe attribute from an object\'s dictionary. For instance, "a.x" has a\nlookup chain starting with "a.__dict__[\'x\']", then\n"type(a).__dict__[\'x\']", and continuing through the base classes of\n"type(a)" excluding metaclasses.\n\nHowever, if the looked-up value is an object defining one of the\ndescriptor methods, then Python may override the default behavior and\ninvoke the descriptor method instead. Where this occurs in the\nprecedence chain depends on which descriptor methods were defined and\nhow they were called. Note that descriptors are only invoked for new\nstyle objects or classes (ones that subclass "object()" or "type()").\n\nThe starting point for descriptor invocation is a binding, "a.x". How\nthe arguments are assembled depends on "a":\n\nDirect Call\n The simplest and least common call is when user code directly\n invokes a descriptor method: "x.__get__(a)".\n\nInstance Binding\n If binding to a new-style object instance, "a.x" is transformed\n into the call: "type(a).__dict__[\'x\'].__get__(a, type(a))".\n\nClass Binding\n If binding to a new-style class, "A.x" is transformed into the\n call: "A.__dict__[\'x\'].__get__(None, A)".\n\nSuper Binding\n If "a" is an instance of "super", then the binding "super(B,\n obj).m()" searches "obj.__class__.__mro__" for the base class "A"\n immediately preceding "B" and then invokes the descriptor with the\n call: "A.__dict__[\'m\'].__get__(obj, obj.__class__)".\n\nFor instance bindings, the precedence of descriptor invocation depends\non the which descriptor methods are defined. A descriptor can define\nany combination of "__get__()", "__set__()" and "__delete__()". If it\ndoes not define "__get__()", then accessing the attribute will return\nthe descriptor object itself unless there is a value in the object\'s\ninstance dictionary. If the descriptor defines "__set__()" and/or\n"__delete__()", it is a data descriptor; if it defines neither, it is\na non-data descriptor. Normally, data descriptors define both\n"__get__()" and "__set__()", while non-data descriptors have just the\n"__get__()" method. Data descriptors with "__set__()" and "__get__()"\ndefined always override a redefinition in an instance dictionary. In\ncontrast, non-data descriptors can be overridden by instances.\n\nPython methods (including "staticmethod()" and "classmethod()") are\nimplemented as non-data descriptors. Accordingly, instances can\nredefine and override methods. This allows individual instances to\nacquire behaviors that differ from other instances of the same class.\n\nThe "property()" function is implemented as a data descriptor.\nAccordingly, instances cannot override the behavior of a property.\n\n\n__slots__\n---------\n\nBy default, instances of both old and new-style classes have a\ndictionary for attribute storage. This wastes space for objects\nhaving very few instance variables. The space consumption can become\nacute when creating large numbers of instances.\n\nThe default can be overridden by defining *__slots__* in a new-style\nclass definition. The *__slots__* declaration takes a sequence of\ninstance variables and reserves just enough space in each instance to\nhold a value for each variable. Space is saved because *__dict__* is\nnot created for each instance.\n\n__slots__\n\n This class variable can be assigned a string, iterable, or sequence\n of strings with variable names used by instances. If defined in a\n new-style class, *__slots__* reserves space for the declared\n variables and prevents the automatic creation of *__dict__* and\n *__weakref__* for each instance.\n\n New in version 2.2.\n\nNotes on using *__slots__*\n\n* When inheriting from a class without *__slots__*, the *__dict__*\n attribute of that class will always be accessible, so a *__slots__*\n definition in the subclass is meaningless.\n\n* Without a *__dict__* variable, instances cannot be assigned new\n variables not listed in the *__slots__* definition. Attempts to\n assign to an unlisted variable name raises "AttributeError". If\n dynamic assignment of new variables is desired, then add\n "\'__dict__\'" to the sequence of strings in the *__slots__*\n declaration.\n\n Changed in version 2.3: Previously, adding "\'__dict__\'" to the\n *__slots__* declaration would not enable the assignment of new\n attributes not specifically listed in the sequence of instance\n variable names.\n\n* Without a *__weakref__* variable for each instance, classes\n defining *__slots__* do not support weak references to its\n instances. If weak reference support is needed, then add\n "\'__weakref__\'" to the sequence of strings in the *__slots__*\n declaration.\n\n Changed in version 2.3: Previously, adding "\'__weakref__\'" to the\n *__slots__* declaration would not enable support for weak\n references.\n\n* *__slots__* are implemented at the class level by creating\n descriptors (Implementing Descriptors) for each variable name. As a\n result, class attributes cannot be used to set default values for\n instance variables defined by *__slots__*; otherwise, the class\n attribute would overwrite the descriptor assignment.\n\n* The action of a *__slots__* declaration is limited to the class\n where it is defined. As a result, subclasses will have a *__dict__*\n unless they also define *__slots__* (which must only contain names\n of any *additional* slots).\n\n* If a class defines a slot also defined in a base class, the\n instance variable defined by the base class slot is inaccessible\n (except by retrieving its descriptor directly from the base class).\n This renders the meaning of the program undefined. In the future, a\n check may be added to prevent this.\n\n* Nonempty *__slots__* does not work for classes derived from\n "variable-length" built-in types such as "long", "str" and "tuple".\n\n* Any non-string iterable may be assigned to *__slots__*. Mappings\n may also be used; however, in the future, special meaning may be\n assigned to the values corresponding to each key.\n\n* *__class__* assignment works only if both classes have the same\n *__slots__*.\n\n Changed in version 2.6: Previously, *__class__* assignment raised an\n error if either new or old class had *__slots__*.\n\n\nCustomizing class creation\n==========================\n\nBy default, new-style classes are constructed using "type()". A class\ndefinition is read into a separate namespace and the value of class\nname is bound to the result of "type(name, bases, dict)".\n\nWhen the class definition is read, if *__metaclass__* is defined then\nthe callable assigned to it will be called instead of "type()". This\nallows classes or functions to be written which monitor or alter the\nclass creation process:\n\n* Modifying the class dictionary prior to the class being created.\n\n* Returning an instance of another class -- essentially performing\n the role of a factory function.\n\nThese steps will have to be performed in the metaclass\'s "__new__()"\nmethod -- "type.__new__()" can then be called from this method to\ncreate a class with different properties. This example adds a new\nelement to the class dictionary before creating the class:\n\n class metacls(type):\n def __new__(mcs, name, bases, dict):\n dict[\'foo\'] = \'metacls was here\'\n return type.__new__(mcs, name, bases, dict)\n\nYou can of course also override other class methods (or add new\nmethods); for example defining a custom "__call__()" method in the\nmetaclass allows custom behavior when the class is called, e.g. not\nalways creating a new instance.\n\n__metaclass__\n\n This variable can be any callable accepting arguments for "name",\n "bases", and "dict". Upon class creation, the callable is used\n instead of the built-in "type()".\n\n New in version 2.2.\n\nThe appropriate metaclass is determined by the following precedence\nrules:\n\n* If "dict[\'__metaclass__\']" exists, it is used.\n\n* Otherwise, if there is at least one base class, its metaclass is\n used (this looks for a *__class__* attribute first and if not found,\n uses its type).\n\n* Otherwise, if a global variable named __metaclass__ exists, it is\n used.\n\n* Otherwise, the old-style, classic metaclass (types.ClassType) is\n used.\n\nThe potential uses for metaclasses are boundless. Some ideas that have\nbeen explored including logging, interface checking, automatic\ndelegation, automatic property creation, proxies, frameworks, and\nautomatic resource locking/synchronization.\n\n\nCustomizing instance and subclass checks\n========================================\n\nNew in version 2.6.\n\nThe following methods are used to override the default behavior of the\n"isinstance()" and "issubclass()" built-in functions.\n\nIn particular, the metaclass "abc.ABCMeta" implements these methods in\norder to allow the addition of Abstract Base Classes (ABCs) as\n"virtual base classes" to any class or type (including built-in\ntypes), including other ABCs.\n\nclass.__instancecheck__(self, instance)\n\n Return true if *instance* should be considered a (direct or\n indirect) instance of *class*. If defined, called to implement\n "isinstance(instance, class)".\n\nclass.__subclasscheck__(self, subclass)\n\n Return true if *subclass* should be considered a (direct or\n indirect) subclass of *class*. If defined, called to implement\n "issubclass(subclass, class)".\n\nNote that these methods are looked up on the type (metaclass) of a\nclass. They cannot be defined as class methods in the actual class.\nThis is consistent with the lookup of special methods that are called\non instances, only in this case the instance is itself a class.\n\nSee also:\n\n **PEP 3119** - Introducing Abstract Base Classes\n Includes the specification for customizing "isinstance()" and\n "issubclass()" behavior through "__instancecheck__()" and\n "__subclasscheck__()", with motivation for this functionality in\n the context of adding Abstract Base Classes (see the "abc"\n module) to the language.\n\n\nEmulating callable objects\n==========================\n\nobject.__call__(self[, args...])\n\n Called when the instance is "called" as a function; if this method\n is defined, "x(arg1, arg2, ...)" is a shorthand for\n "x.__call__(arg1, arg2, ...)".\n\n\nEmulating container types\n=========================\n\nThe following methods can be defined to implement container objects.\nContainers usually are sequences (such as lists or tuples) or mappings\n(like dictionaries), but can represent other containers as well. The\nfirst set of methods is used either to emulate a sequence or to\nemulate a mapping; the difference is that for a sequence, the\nallowable keys should be the integers *k* for which "0 <= k < N" where\n*N* is the length of the sequence, or slice objects, which define a\nrange of items. (For backwards compatibility, the method\n"__getslice__()" (see below) can also be defined to handle simple, but\nnot extended slices.) It is also recommended that mappings provide the\nmethods "keys()", "values()", "items()", "has_key()", "get()",\n"clear()", "setdefault()", "iterkeys()", "itervalues()",\n"iteritems()", "pop()", "popitem()", "copy()", and "update()" behaving\nsimilar to those for Python\'s standard dictionary objects. The\n"UserDict" module provides a "DictMixin" class to help create those\nmethods from a base set of "__getitem__()", "__setitem__()",\n"__delitem__()", and "keys()". Mutable sequences should provide\nmethods "append()", "count()", "index()", "extend()", "insert()",\n"pop()", "remove()", "reverse()" and "sort()", like Python standard\nlist objects. Finally, sequence types should implement addition\n(meaning concatenation) and multiplication (meaning repetition) by\ndefining the methods "__add__()", "__radd__()", "__iadd__()",\n"__mul__()", "__rmul__()" and "__imul__()" described below; they\nshould not define "__coerce__()" or other numerical operators. It is\nrecommended that both mappings and sequences implement the\n"__contains__()" method to allow efficient use of the "in" operator;\nfor mappings, "in" should be equivalent of "has_key()"; for sequences,\nit should search through the values. It is further recommended that\nboth mappings and sequences implement the "__iter__()" method to allow\nefficient iteration through the container; for mappings, "__iter__()"\nshould be the same as "iterkeys()"; for sequences, it should iterate\nthrough the values.\n\nobject.__len__(self)\n\n Called to implement the built-in function "len()". Should return\n the length of the object, an integer ">=" 0. Also, an object that\n doesn\'t define a "__nonzero__()" method and whose "__len__()"\n method returns zero is considered to be false in a Boolean context.\n\nobject.__getitem__(self, key)\n\n Called to implement evaluation of "self[key]". For sequence types,\n the accepted keys should be integers and slice objects. Note that\n the special interpretation of negative indexes (if the class wishes\n to emulate a sequence type) is up to the "__getitem__()" method. If\n *key* is of an inappropriate type, "TypeError" may be raised; if of\n a value outside the set of indexes for the sequence (after any\n special interpretation of negative values), "IndexError" should be\n raised. For mapping types, if *key* is missing (not in the\n container), "KeyError" should be raised.\n\n Note: "for" loops expect that an "IndexError" will be raised for\n illegal indexes to allow proper detection of the end of the\n sequence.\n\nobject.__missing__(self, key)\n\n Called by "dict"."__getitem__()" to implement "self[key]" for dict\n subclasses when key is not in the dictionary.\n\nobject.__setitem__(self, key, value)\n\n Called to implement assignment to "self[key]". Same note as for\n "__getitem__()". This should only be implemented for mappings if\n the objects support changes to the values for keys, or if new keys\n can be added, or for sequences if elements can be replaced. The\n same exceptions should be raised for improper *key* values as for\n the "__getitem__()" method.\n\nobject.__delitem__(self, key)\n\n Called to implement deletion of "self[key]". Same note as for\n "__getitem__()". This should only be implemented for mappings if\n the objects support removal of keys, or for sequences if elements\n can be removed from the sequence. The same exceptions should be\n raised for improper *key* values as for the "__getitem__()" method.\n\nobject.__iter__(self)\n\n This method is called when an iterator is required for a container.\n This method should return a new iterator object that can iterate\n over all the objects in the container. For mappings, it should\n iterate over the keys of the container, and should also be made\n available as the method "iterkeys()".\n\n Iterator objects also need to implement this method; they are\n required to return themselves. For more information on iterator\n objects, see Iterator Types.\n\nobject.__reversed__(self)\n\n Called (if present) by the "reversed()" built-in to implement\n reverse iteration. It should return a new iterator object that\n iterates over all the objects in the container in reverse order.\n\n If the "__reversed__()" method is not provided, the "reversed()"\n built-in will fall back to using the sequence protocol ("__len__()"\n and "__getitem__()"). Objects that support the sequence protocol\n should only provide "__reversed__()" if they can provide an\n implementation that is more efficient than the one provided by\n "reversed()".\n\n New in version 2.6.\n\nThe membership test operators ("in" and "not in") are normally\nimplemented as an iteration through a sequence. However, container\nobjects can supply the following special method with a more efficient\nimplementation, which also does not require the object be a sequence.\n\nobject.__contains__(self, item)\n\n Called to implement membership test operators. Should return true\n if *item* is in *self*, false otherwise. For mapping objects, this\n should consider the keys of the mapping rather than the values or\n the key-item pairs.\n\n For objects that don\'t define "__contains__()", the membership test\n first tries iteration via "__iter__()", then the old sequence\n iteration protocol via "__getitem__()", see this section in the\n language reference.\n\n\nAdditional methods for emulation of sequence types\n==================================================\n\nThe following optional methods can be defined to further emulate\nsequence objects. Immutable sequences methods should at most only\ndefine "__getslice__()"; mutable sequences might define all three\nmethods.\n\nobject.__getslice__(self, i, j)\n\n Deprecated since version 2.0: Support slice objects as parameters\n to the "__getitem__()" method. (However, built-in types in CPython\n currently still implement "__getslice__()". Therefore, you have to\n override it in derived classes when implementing slicing.)\n\n Called to implement evaluation of "self[i:j]". The returned object\n should be of the same type as *self*. Note that missing *i* or *j*\n in the slice expression are replaced by zero or "sys.maxsize",\n respectively. If negative indexes are used in the slice, the\n length of the sequence is added to that index. If the instance does\n not implement the "__len__()" method, an "AttributeError" is\n raised. No guarantee is made that indexes adjusted this way are not\n still negative. Indexes which are greater than the length of the\n sequence are not modified. If no "__getslice__()" is found, a slice\n object is created instead, and passed to "__getitem__()" instead.\n\nobject.__setslice__(self, i, j, sequence)\n\n Called to implement assignment to "self[i:j]". Same notes for *i*\n and *j* as for "__getslice__()".\n\n This method is deprecated. If no "__setslice__()" is found, or for\n extended slicing of the form "self[i:j:k]", a slice object is\n created, and passed to "__setitem__()", instead of "__setslice__()"\n being called.\n\nobject.__delslice__(self, i, j)\n\n Called to implement deletion of "self[i:j]". Same notes for *i* and\n *j* as for "__getslice__()". This method is deprecated. If no\n "__delslice__()" is found, or for extended slicing of the form\n "self[i:j:k]", a slice object is created, and passed to\n "__delitem__()", instead of "__delslice__()" being called.\n\nNotice that these methods are only invoked when a single slice with a\nsingle colon is used, and the slice method is available. For slice\noperations involving extended slice notation, or in absence of the\nslice methods, "__getitem__()", "__setitem__()" or "__delitem__()" is\ncalled with a slice object as argument.\n\nThe following example demonstrate how to make your program or module\ncompatible with earlier versions of Python (assuming that methods\n"__getitem__()", "__setitem__()" and "__delitem__()" support slice\nobjects as arguments):\n\n class MyClass:\n ...\n def __getitem__(self, index):\n ...\n def __setitem__(self, index, value):\n ...\n def __delitem__(self, index):\n ...\n\n if sys.version_info < (2, 0):\n # They won\'t be defined if version is at least 2.0 final\n\n def __getslice__(self, i, j):\n return self[max(0, i):max(0, j):]\n def __setslice__(self, i, j, seq):\n self[max(0, i):max(0, j):] = seq\n def __delslice__(self, i, j):\n del self[max(0, i):max(0, j):]\n ...\n\nNote the calls to "max()"; these are necessary because of the handling\nof negative indices before the "__*slice__()" methods are called.\nWhen negative indexes are used, the "__*item__()" methods receive them\nas provided, but the "__*slice__()" methods get a "cooked" form of the\nindex values. For each negative index value, the length of the\nsequence is added to the index before calling the method (which may\nstill result in a negative index); this is the customary handling of\nnegative indexes by the built-in sequence types, and the "__*item__()"\nmethods are expected to do this as well. However, since they should\nalready be doing that, negative indexes cannot be passed in; they must\nbe constrained to the bounds of the sequence before being passed to\nthe "__*item__()" methods. Calling "max(0, i)" conveniently returns\nthe proper value.\n\n\nEmulating numeric types\n=======================\n\nThe following methods can be defined to emulate numeric objects.\nMethods corresponding to operations that are not supported by the\nparticular kind of number implemented (e.g., bitwise operations for\nnon-integral numbers) should be left undefined.\n\nobject.__add__(self, other)\nobject.__sub__(self, other)\nobject.__mul__(self, other)\nobject.__floordiv__(self, other)\nobject.__mod__(self, other)\nobject.__divmod__(self, other)\nobject.__pow__(self, other[, modulo])\nobject.__lshift__(self, other)\nobject.__rshift__(self, other)\nobject.__and__(self, other)\nobject.__xor__(self, other)\nobject.__or__(self, other)\n\n These methods are called to implement the binary arithmetic\n operations ("+", "-", "*", "//", "%", "divmod()", "pow()", "**",\n "<<", ">>", "&", "^", "|"). For instance, to evaluate the\n expression "x + y", where *x* is an instance of a class that has an\n "__add__()" method, "x.__add__(y)" is called. The "__divmod__()"\n method should be the equivalent to using "__floordiv__()" and\n "__mod__()"; it should not be related to "__truediv__()" (described\n below). Note that "__pow__()" should be defined to accept an\n optional third argument if the ternary version of the built-in\n "pow()" function is to be supported.\n\n If one of those methods does not support the operation with the\n supplied arguments, it should return "NotImplemented".\n\nobject.__div__(self, other)\nobject.__truediv__(self, other)\n\n The division operator ("/") is implemented by these methods. The\n "__truediv__()" method is used when "__future__.division" is in\n effect, otherwise "__div__()" is used. If only one of these two\n methods is defined, the object will not support division in the\n alternate context; "TypeError" will be raised instead.\n\nobject.__radd__(self, other)\nobject.__rsub__(self, other)\nobject.__rmul__(self, other)\nobject.__rdiv__(self, other)\nobject.__rtruediv__(self, other)\nobject.__rfloordiv__(self, other)\nobject.__rmod__(self, other)\nobject.__rdivmod__(self, other)\nobject.__rpow__(self, other)\nobject.__rlshift__(self, other)\nobject.__rrshift__(self, other)\nobject.__rand__(self, other)\nobject.__rxor__(self, other)\nobject.__ror__(self, other)\n\n These methods are called to implement the binary arithmetic\n operations ("+", "-", "*", "/", "%", "divmod()", "pow()", "**",\n "<<", ">>", "&", "^", "|") with reflected (swapped) operands.\n These functions are only called if the left operand does not\n support the corresponding operation and the operands are of\n different types. [2] For instance, to evaluate the expression "x -\n y", where *y* is an instance of a class that has an "__rsub__()"\n method, "y.__rsub__(x)" is called if "x.__sub__(y)" returns\n *NotImplemented*.\n\n Note that ternary "pow()" will not try calling "__rpow__()" (the\n coercion rules would become too complicated).\n\n Note: If the right operand\'s type is a subclass of the left\n operand\'s type and that subclass provides the reflected method\n for the operation, this method will be called before the left\n operand\'s non-reflected method. This behavior allows subclasses\n to override their ancestors\' operations.\n\nobject.__iadd__(self, other)\nobject.__isub__(self, other)\nobject.__imul__(self, other)\nobject.__idiv__(self, other)\nobject.__itruediv__(self, other)\nobject.__ifloordiv__(self, other)\nobject.__imod__(self, other)\nobject.__ipow__(self, other[, modulo])\nobject.__ilshift__(self, other)\nobject.__irshift__(self, other)\nobject.__iand__(self, other)\nobject.__ixor__(self, other)\nobject.__ior__(self, other)\n\n These methods are called to implement the augmented arithmetic\n assignments ("+=", "-=", "*=", "/=", "//=", "%=", "**=", "<<=",\n ">>=", "&=", "^=", "|="). These methods should attempt to do the\n operation in-place (modifying *self*) and return the result (which\n could be, but does not have to be, *self*). If a specific method\n is not defined, the augmented assignment falls back to the normal\n methods. For instance, to execute the statement "x += y", where\n *x* is an instance of a class that has an "__iadd__()" method,\n "x.__iadd__(y)" is called. If *x* is an instance of a class that\n does not define a "__iadd__()" method, "x.__add__(y)" and\n "y.__radd__(x)" are considered, as with the evaluation of "x + y".\n\nobject.__neg__(self)\nobject.__pos__(self)\nobject.__abs__(self)\nobject.__invert__(self)\n\n Called to implement the unary arithmetic operations ("-", "+",\n "abs()" and "~").\n\nobject.__complex__(self)\nobject.__int__(self)\nobject.__long__(self)\nobject.__float__(self)\n\n Called to implement the built-in functions "complex()", "int()",\n "long()", and "float()". Should return a value of the appropriate\n type.\n\nobject.__oct__(self)\nobject.__hex__(self)\n\n Called to implement the built-in functions "oct()" and "hex()".\n Should return a string value.\n\nobject.__index__(self)\n\n Called to implement "operator.index()". Also called whenever\n Python needs an integer object (such as in slicing). Must return\n an integer (int or long).\n\n New in version 2.5.\n\nobject.__coerce__(self, other)\n\n Called to implement "mixed-mode" numeric arithmetic. Should either\n return a 2-tuple containing *self* and *other* converted to a\n common numeric type, or "None" if conversion is impossible. When\n the common type would be the type of "other", it is sufficient to\n return "None", since the interpreter will also ask the other object\n to attempt a coercion (but sometimes, if the implementation of the\n other type cannot be changed, it is useful to do the conversion to\n the other type here). A return value of "NotImplemented" is\n equivalent to returning "None".\n\n\nCoercion rules\n==============\n\nThis section used to document the rules for coercion. As the language\nhas evolved, the coercion rules have become hard to document\nprecisely; documenting what one version of one particular\nimplementation does is undesirable. Instead, here are some informal\nguidelines regarding coercion. In Python 3, coercion will not be\nsupported.\n\n* If the left operand of a % operator is a string or Unicode object,\n no coercion takes place and the string formatting operation is\n invoked instead.\n\n* It is no longer recommended to define a coercion operation. Mixed-\n mode operations on types that don\'t define coercion pass the\n original arguments to the operation.\n\n* New-style classes (those derived from "object") never invoke the\n "__coerce__()" method in response to a binary operator; the only\n time "__coerce__()" is invoked is when the built-in function\n "coerce()" is called.\n\n* For most intents and purposes, an operator that returns\n "NotImplemented" is treated the same as one that is not implemented\n at all.\n\n* Below, "__op__()" and "__rop__()" are used to signify the generic\n method names corresponding to an operator; "__iop__()" is used for\n the corresponding in-place operator. For example, for the operator\n \'"+"\', "__add__()" and "__radd__()" are used for the left and right\n variant of the binary operator, and "__iadd__()" for the in-place\n variant.\n\n* For objects *x* and *y*, first "x.__op__(y)" is tried. If this is\n not implemented or returns "NotImplemented", "y.__rop__(x)" is\n tried. If this is also not implemented or returns "NotImplemented",\n a "TypeError" exception is raised. But see the following exception:\n\n* Exception to the previous item: if the left operand is an instance\n of a built-in type or a new-style class, and the right operand is an\n instance of a proper subclass of that type or class and overrides\n the base\'s "__rop__()" method, the right operand\'s "__rop__()"\n method is tried *before* the left operand\'s "__op__()" method.\n\n This is done so that a subclass can completely override binary\n operators. Otherwise, the left operand\'s "__op__()" method would\n always accept the right operand: when an instance of a given class\n is expected, an instance of a subclass of that class is always\n acceptable.\n\n* When either operand type defines a coercion, this coercion is\n called before that type\'s "__op__()" or "__rop__()" method is\n called, but no sooner. If the coercion returns an object of a\n different type for the operand whose coercion is invoked, part of\n the process is redone using the new object.\n\n* When an in-place operator (like \'"+="\') is used, if the left\n operand implements "__iop__()", it is invoked without any coercion.\n When the operation falls back to "__op__()" and/or "__rop__()", the\n normal coercion rules apply.\n\n* In "x + y", if *x* is a sequence that implements sequence\n concatenation, sequence concatenation is invoked.\n\n* In "x * y", if one operand is a sequence that implements sequence\n repetition, and the other is an integer ("int" or "long"), sequence\n repetition is invoked.\n\n* Rich comparisons (implemented by methods "__eq__()" and so on)\n never use coercion. Three-way comparison (implemented by\n "__cmp__()") does use coercion under the same conditions as other\n binary operations use it.\n\n* In the current implementation, the built-in numeric types "int",\n "long", "float", and "complex" do not use coercion. All these types\n implement a "__coerce__()" method, for use by the built-in\n "coerce()" function.\n\n Changed in version 2.7: The complex type no longer makes implicit\n calls to the "__coerce__()" method for mixed-type binary arithmetic\n operations.\n\n\nWith Statement Context Managers\n===============================\n\nNew in version 2.5.\n\nA *context manager* is an object that defines the runtime context to\nbe established when executing a "with" statement. The context manager\nhandles the entry into, and the exit from, the desired runtime context\nfor the execution of the block of code. Context managers are normally\ninvoked using the "with" statement (described in section The with\nstatement), but can also be used by directly invoking their methods.\n\nTypical uses of context managers include saving and restoring various\nkinds of global state, locking and unlocking resources, closing opened\nfiles, etc.\n\nFor more information on context managers, see Context Manager Types.\n\nobject.__enter__(self)\n\n Enter the runtime context related to this object. The "with"\n statement will bind this method\'s return value to the target(s)\n specified in the "as" clause of the statement, if any.\n\nobject.__exit__(self, exc_type, exc_value, traceback)\n\n Exit the runtime context related to this object. The parameters\n describe the exception that caused the context to be exited. If the\n context was exited without an exception, all three arguments will\n be "None".\n\n If an exception is supplied, and the method wishes to suppress the\n exception (i.e., prevent it from being propagated), it should\n return a true value. Otherwise, the exception will be processed\n normally upon exit from this method.\n\n Note that "__exit__()" methods should not reraise the passed-in\n exception; this is the caller\'s responsibility.\n\nSee also:\n\n **PEP 343** - The "with" statement\n The specification, background, and examples for the Python "with"\n statement.\n\n\nSpecial method lookup for old-style classes\n===========================================\n\nFor old-style classes, special methods are always looked up in exactly\nthe same way as any other method or attribute. This is the case\nregardless of whether the method is being looked up explicitly as in\n"x.__getitem__(i)" or implicitly as in "x[i]".\n\nThis behaviour means that special methods may exhibit different\nbehaviour for different instances of a single old-style class if the\nappropriate special attributes are set differently:\n\n >>> class C:\n ... pass\n ...\n >>> c1 = C()\n >>> c2 = C()\n >>> c1.__len__ = lambda: 5\n >>> c2.__len__ = lambda: 9\n >>> len(c1)\n 5\n >>> len(c2)\n 9\n\n\nSpecial method lookup for new-style classes\n===========================================\n\nFor new-style classes, implicit invocations of special methods are\nonly guaranteed to work correctly if defined on an object\'s type, not\nin the object\'s instance dictionary. That behaviour is the reason why\nthe following code raises an exception (unlike the equivalent example\nwith old-style classes):\n\n >>> class C(object):\n ... pass\n ...\n >>> c = C()\n >>> c.__len__ = lambda: 5\n >>> len(c)\n Traceback (most recent call last):\n File "<stdin>", line 1, in <module>\n TypeError: object of type \'C\' has no len()\n\nThe rationale behind this behaviour lies with a number of special\nmethods such as "__hash__()" and "__repr__()" that are implemented by\nall objects, including type objects. If the implicit lookup of these\nmethods used the conventional lookup process, they would fail when\ninvoked on the type object itself:\n\n >>> 1 .__hash__() == hash(1)\n True\n >>> int.__hash__() == hash(int)\n Traceback (most recent call last):\n File "<stdin>", line 1, in <module>\n TypeError: descriptor \'__hash__\' of \'int\' object needs an argument\n\nIncorrectly attempting to invoke an unbound method of a class in this\nway is sometimes referred to as \'metaclass confusion\', and is avoided\nby bypassing the instance when looking up special methods:\n\n >>> type(1).__hash__(1) == hash(1)\n True\n >>> type(int).__hash__(int) == hash(int)\n True\n\nIn addition to bypassing any instance attributes in the interest of\ncorrectness, implicit special method lookup generally also bypasses\nthe "__getattribute__()" method even of the object\'s metaclass:\n\n >>> class Meta(type):\n ... def __getattribute__(*args):\n ... print "Metaclass getattribute invoked"\n ... return type.__getattribute__(*args)\n ...\n >>> class C(object):\n ... __metaclass__ = Meta\n ... def __len__(self):\n ... return 10\n ... def __getattribute__(*args):\n ... print "Class getattribute invoked"\n ... return object.__getattribute__(*args)\n ...\n >>> c = C()\n >>> c.__len__() # Explicit lookup via instance\n Class getattribute invoked\n 10\n >>> type(c).__len__(c) # Explicit lookup via type\n Metaclass getattribute invoked\n 10\n >>> len(c) # Implicit lookup\n 10\n\nBypassing the "__getattribute__()" machinery in this fashion provides\nsignificant scope for speed optimisations within the interpreter, at\nthe cost of some flexibility in the handling of special methods (the\nspecial method *must* be set on the class object itself in order to be\nconsistently invoked by the interpreter).\n\n-[ Footnotes ]-\n\n[1] It *is* possible in some cases to change an object\'s type,\n under certain controlled conditions. It generally isn\'t a good\n idea though, since it can lead to some very strange behaviour if\n it is handled incorrectly.\n\n[2] For operands of the same type, it is assumed that if the non-\n reflected method (such as "__add__()") fails the operation is not\n supported, which is why the reflected method is not called.\n',
|